U.S. patent application number 11/541282 was filed with the patent office on 2008-04-03 for method and apparatus for active noise cancellation.
This patent application is currently assigned to SRI International. Invention is credited to Venkata Ramana Rao Gadde, Martin Graciarena, Anand Venkataraman.
Application Number | 20080082326 11/541282 |
Document ID | / |
Family ID | 39262072 |
Filed Date | 2008-04-03 |
United States Patent
Application |
20080082326 |
Kind Code |
A1 |
Venkataraman; Anand ; et
al. |
April 3, 2008 |
Method and apparatus for active noise cancellation
Abstract
In one embodiment, the present invention is a method and
apparatus for active noise cancellation. In one embodiment, a
method for recognizing user speech in an audio signal received by a
media system (where the audio signal includes user speech and
ambient audio output produced by the media system and/or other
devices) includes canceling portions of the audio signal associated
with the ambient audio output and applying speech recognition
processing to an uncancelled remainder of the audio signal.
Inventors: |
Venkataraman; Anand; (Palo
Alto, CA) ; Gadde; Venkata Ramana Rao; (Santa Clara,
CA) ; Graciarena; Martin; (Menlo Park, CA) |
Correspondence
Address: |
PATTERSON & SHERIDAN L.L.P.
595 SHREWSBURY AVE, STE 100, FIRST FLOOR
SHREWSBURY
NJ
07702
US
|
Assignee: |
SRI International
|
Family ID: |
39262072 |
Appl. No.: |
11/541282 |
Filed: |
September 28, 2006 |
Current U.S.
Class: |
704/228 |
Current CPC
Class: |
G10L 15/20 20130101;
G10L 2015/088 20130101; G10L 21/02 20130101 |
Class at
Publication: |
704/228 |
International
Class: |
G10L 21/02 20060101
G10L021/02 |
Claims
1. A method for recognizing user speech in an audio signal received
by a media system, said audio signal comprising user speech and
ambient audio output, the method comprising: canceling portions of
said audio signal associated with said ambient audio output; and
applying speech recognition processing to an uncancelled remainder
of said audio signal.
2. The method of claim 1, wherein said canceling comprises:
applying at least one offset to said audio signal, said at least
one offset compensating for an audio signal produced by at least
one output channel of said media system.
3. The method of claim 2, wherein said applying comprises:
calculating said at least one offset in accordance with a response
of said at least one output channel to a calibration signal.
4. The method of claim 1, wherein said applying comprises:
detecting a trigger word in said audio signal; and commencing
speech recognition processing of at least a portion of said audio
signal following said trigger word.
5. The method of claim 4, wherein said detecting comprises:
scanning said audio signal for one or more trigger words including
said trigger word on a substantially continuous basis.
6. The method of claim 4, wherein said at least a portion of said
audio signal following said trigger word comprises a command
uttered by said user.
7. The method of claim 6, further comprising: taking an action in
accordance with said command.
8. The method of claim 1, wherein said audio signal is received via
a microphone interfaced to said media system.
9. The method of claim 1, wherein said canceling comprises:
determining a level of distortion in said audio signal that is
attributable to said ambient audio output being received in said
audio signal; and compensating for said level of distortion prior
to applying said speech recognition processing.
10. The method of claim 1, wherein said media system comprises at
least one of: a personal computer media system, a television, a car
audio system and a car navigation system.
11. A computer readable medium containing an executable program for
recognizing user speech in an audio signal received by a media
system, said audio signal comprising user speech and ambient audio
output, where the program performs the steps of: canceling portions
of said audio signal associated with said ambient audio output; and
applying speech recognition processing to an uncancelled remainder
of said audio signal.
12. The computer readable medium of claim 11, wherein said
canceling comprises: applying at least one offset to said audio
signal, said at least one offset compensating for an audio signal
produced by at least one output channel of said media system.
13. The computer readable medium of claim 12, wherein said applying
comprises: calculating said at least one offset in accordance with
a response of said at least one output channel to a calibration
signal.
14. The computer readable medium of claim 11, wherein said applying
comprises: detecting a trigger word in said audio signal; and
commencing speech recognition processing of at least a portion of
said audio signal following said trigger word.
15. The computer readable medium of claim 14, wherein said
detecting comprises: scanning said audio signal for one or more
trigger words including said trigger word on a substantially
continuous basis.
16. The computer readable medium of claim 14, wherein said at least
a portion of said audio signal following said trigger word
comprises a command uttered by said user.
17. The computer readable medium of claim 16, further comprising:
taking an action in accordance with said command.
18. The computer readable medium of claim 11, wherein said audio
signal is received via a microphone interfaced to said media
system.
19. The computer readable medium of claim 11, wherein said
canceling comprises: determining a level of distortion in said
audio signal that is attributable to said ambient audio output
being received in said audio signal; and compensating for said
level of distortion prior to applying said speech recognition
processing.
20. The computer readable medium of claim 11, wherein said media
system comprises at least one of: a personal computer media system,
a television, a car audio system and a car navigation system.
21. An apparatus for recognizing user speech in an audio signal
received by a media system, said audio signal comprising user
speech and ambient audio output, comprising: means for canceling
portions of said audio signal associated with said ambient audio
output; and means for applying speech recognition processing to an
uncancelled remainder of said audio signal.
22. The apparatus medium of claim 21, wherein said means for
canceling comprises: means for applying at least one offset to said
audio signal, said at least one offset compensating for an audio
signal produced by at least one output channel of said media
system.
23. The apparatus medium of claim 22, wherein said means for
applying comprises: means for calculating said at least one offset
in accordance with a response of said at least one output channel
to a calibration signal.
24. The apparatus medium of claim 21, wherein said means for
applying comprises: means for detecting a trigger word in said
audio signal; and means for commencing speech recognition
processing of at least a portion of said audio signal following
said trigger word.
25. The apparatus of claim 24, wherein said means for detecting
scans said audio signal for one or more trigger words including
said trigger word on a substantially continuous basis.
26. The apparatus of claim 24, wherein said at least a portion of
said audio signal following said trigger word comprises a command
uttered by said user.
27. The apparatus of claim 26, further comprising: means for taking
an action in accordance with said command.
28. The apparatus of claim 21, wherein further comprising a
microphone interfaced to said media system for receiving said audio
signal.
29. The apparatus of claim 21, wherein said means for canceling
comprises: means for determining a level of distortion in said
audio signal that is attributable to said ambient audio output
being received in said audio signal; and means for compensating for
said level of distortion prior to applying said speech recognition
processing.
30. The apparatus of claim 21, wherein said media system comprises
at least one of: a personal computer media system, a television, a
car audio system and a car navigation system.
Description
FIELD OF THE INVENTION
[0001] The present invention relates generally to speech
recognition and relates more particularly to speech recognition in
noisy environments.
BACKGROUND OF THE INVENTION
[0002] Presently, remote control of media systems, including media
center applications such as channel guide or jukebox applications
and car audio systems is difficult. In the case of media center
applications, the applications are typically controlled by using a
mouse or by issuing a voice command. In the case of voice command,
however, ambient noise (such as that produced by the media center
application itself) often makes it difficult for speech recognition
software to successfully recognize the issued commands.
[0003] Thus, there is a need in the art for a method and apparatus
for active noise cancellation (i.e., cancellation of noise produced
by a media system itself).
SUMMARY OF THE INVENTION
[0004] In one embodiment, the present invention is a method and
apparatus for active noise cancellation. In one embodiment, a
method for recognizing user speech in an audio signal received by a
media system (where the audio signal includes user speech and
ambient audio output produced by the media system and/or other
devices) includes canceling portions of the audio signal associated
with the ambient audio output and applying speech recognition
processing to an uncancelled remainder of the audio signal.
BRIEF DESCRIPTION OF THE DRAWINGS
[0005] The teaching of the present invention can be readily
understood by considering the following detailed description in
conjunction with the accompanying drawings, in which:
[0006] FIG. 1 is a flow diagram illustrating one embodiment of a
method for active noise cancellation, according to the present
invention;
[0007] FIG. 2 is a flow diagram illustrating one embodiment of a
method for calibrating an audio output system for active noise
cancellation, according to the present invention; and
[0008] FIG. 3 is a high level block diagram of the noise
cancellation method that is implemented using a general purpose
computing device.
[0009] To facilitate understanding, identical reference numerals
have been used, where possible, to designate identical elements
that are common to the figures.
DETAILED DESCRIPTION
[0010] In one embodiment, the present invention relates to active
noise cancellation for speech recognition applications, such as
speech recognition applications used to control media systems
(e.g., systems that at least produce audio output, and may produce
other outputs such as video) including media center applications
running on personal computers (PCs), televisions and car audio and
navigation systems. Embodiments of the invention exploit the fact
that a media system has knowledge of the audio signals being
delivered via its output channels. This knowledge may be applied to
cancel out ambient noise produced by the media system in audio
signals received (e.g., via a microphone) by a speech recognition
application running on the media system. The accuracy of subsequent
speech recognition processing of the received audio signals (e.g.,
to extract spoken user commands) is thus significantly
enhanced.
[0011] FIG. 1 is a flow diagram illustrating one embodiment of a
method 100 for active noise cancellation, according to the present
invention. The method 100 may be implemented, for example, on a
personal computer that runs a media center application or in a car
audio system. The method 100 is initialized at step 102 and
proceeds to step 104, where the method 100 receives an audio signal
that originates external to the media system. In one embodiment,
the received audio signal comprises at least audio output produced
by the media system. In a further embodiment, the received audio
signal also comprises user speech (e.g., a spoken user command)
and/or other ambient noise. In one embodiment, the audio signal is
received via a microphone that is interfaced to the system. In one
embodiment, the microphone is incorporated in at least one of: a
remote control, an amplifier or a media center device (e.g., PC,
television, stereo, etc.) or component thereof.
[0012] In step 106, the method 100 cancels portions of the received
audio signal that are associated with output channels of the media
system. Thus, for example, if the media system is a jukebox media
center application that emits six-channel audio from a PC (e.g.,
amplified and fed through six speakers placed at various locations
within a room including the PC), the six channels of emitted audio
are precisely the signals that need to be removed from the system's
microphone input. In one embodiment, the portions of the received
audio signal that are associated with output channels of the media
system are cancelled by subtracting those portions of the audio
signal from the received audio signal. In one embodiment, this is
done by applying offsets for each output channel, where the offsets
are calculated in accordance with a previously applied calibration
technique described in greater detail with reference to FIG. 2.
[0013] In step 108, the method 100 scans the received audio signal
for a trigger word. The trigger word is a word that indicates that
a user of the media center application is issuing a voice command,
and may be followed by the voice command. For example, the user may
utter the phrase "<TRIGGER_WORD> Switch to KQED", where
"switch to KQED" is the command that the user wants the media
center application to execute. The utterance of a trigger word
triggers speech recognition in the media center application.
[0014] In step 110, the method 100 determines whether a trigger
word has been detected in the received audio signal. If the method
100 concludes in step 110 that a trigger word has not been
detected, the method 100 returns to step 104 and proceeds as
described above to continue to process the audio signal and scan
for trigger words.
[0015] Alternatively, if the method 100 concludes in step 110 that
a trigger word has been detected, the method 100 proceeds to step
112 and applies speech recognition processing to the incoming audio
signal in order to extract the voice command (e.g., following the
trigger word). In one embodiment, the speech recognition
application processes the audio signal in accordance with a small
and tight speech recognition grammar.
[0016] In step 114, the method 100 takes some action in accordance
with the extracted command (e.g., changes a radio station to KQED
in the case of the example above). The method 100 then returns to
step 104 and proceeds as described above to continue to process the
audio signal and scan for trigger words.
[0017] By applying knowledge of the audio signals produced by the
media system to cancel ambient noise in the received audio signal,
more accurate recognition of spoken user commands can be achieved.
That is, the signals associated with the media system's output
channels can be removed from the received audio signal (e.g., as
picked up by a microphone) in a fairly precise manner. Thus, even a
user command that is spoken softly and/or from a distance away can
be detected and recognized, despite the ambient noise produced by
the media system.
[0018] FIG. 2 is a flow diagram illustrating one embodiment of a
method 200 for calibrating an audio output system for active noise
cancellation, according to the present invention. That is, the
method 200 determines the proper attenuations and offsets to be
applied to a received audio signal in order to cancel ambient noise
produced, for example, by a media system. The method 200 may thus
be applied prior to execution of the method 100, so that ambient
noise produced by the media system can be cancelled in the received
audio signal.
[0019] The method 200 is initialized at step 202 and proceeds to
step 204, where the method 200 selects and activates one audio
output channel. The selected channel is a channel that has not yet
been calibrated.
[0020] In step 206, the method 200 sends a calibration signal to
the activated output channel. The method 200 then proceeds to step
208 and measures the channel's response to the calibration signal,
e.g., as determined by reception at a microphone interfaced to the
media system. In one embodiment, the response includes the time
elapsed between the sending of the calibration signal and the
reception of the channel's response, as well as the distortion in
the channel's response (i.e., caused by the channel's audio output
being emitted via the channel and then picked up again by the
microphone).
[0021] In step 210, the method 200, calculates, in accordance with
the response measured in step 208, the attenuation (e.g., to
compensate for distortions) and offsets for the activated channel.
The calculated offsets are the offsets that will later be applied
to cancel the output from the activated channel in an audio signal
received by the media system (e.g., as described with respect to
step 106 of FIG. 1).
[0022] In step 212, the method 200 determines whether there are any
output channels that remain to be calibrated. If the method 200
concludes in step 212 that at least one output channel still
requires calibration, the method 200 returns to step 204 and
proceeds as described above in order to calibrate the remaining
channel(s). Alternatively, if the method 200 concludes in step 212
that there are no uncalibrated output channels remaining, the
method 200 terminates in step 214.
[0023] FIG. 3 is a high level block diagram of the noise
cancellation method that is implemented using a general purpose
computing device 300. In one embodiment, a general purpose
computing device 300 comprises a processor 302, a memory 304, a
noise cancellation module 305 and various input/output (I/O)
devices 306 such as a display, a keyboard, a mouse, a modem, and
the like. In one embodiment, at least one I/O device is a storage
device (e.g., a disk drive, an optical disk drive, a floppy disk
drive). It should be understood that the noise cancellation module
305 can be implemented as a physical device or subsystem that is
coupled to a processor through a communication channel.
[0024] Alternatively, the noise cancellation module 305 can be
represented by one or more software applications (or even a
combination of software and hardware, e.g., using Application
Specific Integrated Circuits (ASIC)), where the software is loaded
from a storage medium (e.g., I/O devices 306) and operated by the
processor 302 in the memory 304 of the general purpose computing
device 300. Thus, in one embodiment, the noise cancellation module
305 for canceling ambient noise in speech recognition applications
described herein with reference to the preceding Figures can be
stored on a computer readable medium or carrier (e.g., RAM,
magnetic or optical drive or diskette, and the like).
[0025] Those skilled in the art will appreciate that the concepts
of the present invention may be advantageously deployed in a
variety of applications, and not just those running on media center
PCs. For instance, any audio application in which speech-driven
control is desirable and the audio output is knowable may benefit
from application of the present invention, including car audio
systems and the like. The present invention may also aid users of
telephones, including cellular phones, particularly when using a
telephone in a noisy environment such as in an automobile (in such
a case, the cellular phone's communicative coupling to the audio
source may comprise, for example, a wireless personal or local area
network such as a Bluetooth connection, a WiFi connection or a
built-in wire).
[0026] In addition, the present invention may be advantageously
deployed to control a variety of other (non-media center) PC
applications, such as dictation programs, Voice over IP (VoIP)
applications, and other applications that are compatible with voice
control.
[0027] Thus, the present invention represents a significant
advancement in the field of speech recognition applications.
Embodiments of the invention exploit the fact that a media system
has knowledge of the audio signals being delivered via its output
channels. This knowledge may be applied to cancel out ambient noise
produced by the media system in audio signals received (e.g., via a
microphone) by a speech recognition application running on the PC.
The accuracy of subsequent speech recognition processing of the
received audio signals (e.g., to extract spoken user commands) is
thus significantly enhanced.
[0028] While various embodiments have been described above, it
should be understood that they have been presented by way of
example only, and not limitation. Thus, the breadth and scope of a
preferred embodiment should not be limited by any of the
above-described exemplary embodiments, but should be defined only
in accordance with the following claims and their equivalents.
* * * * *